Skip to content

Latest commit

 

History

History
1789 lines (1366 loc) · 72.7 KB

3.11.rst

File metadata and controls

1789 lines (1366 loc) · 72.7 KB
Raw

What's New In Python 3.11

Release
Date

This article explains the new features in Python 3.11, compared to 3.10.

For full details, see the changelog <changelog>.

Note

Prerelease users should be aware that this document is currently in draft form. It will be updated substantially as Python 3.11 moves towards release, so it's worth checking back even after reading earlier versions.

Summary -- Release highlights

  • Python 3.11 is up to 10-60% faster than Python 3.10. On average, we measured a 1.22x speedup on the standard benchmark suite. See Faster CPython for details.

New syntax features:

  • 654: Exception Groups and except*. (Contributed by Irit Katriel in 45292.)

New typing features:

  • 646: Variadic generics.
  • 655: Marking individual TypedDict items as required or potentially-missing.
  • 673: Self type.
  • 675: Arbitrary literal string type.

New Features

Enhanced error locations in tracebacks

When printing tracebacks, the interpreter will now point to the exact expression that caused the error instead of just the line. For example:

Traceback (most recent call last):
  File "distance.py", line 11, in <module>
    print(manhattan_distance(p1, p2))
          ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "distance.py", line 6, in manhattan_distance
    return abs(point_1.x - point_2.x) + abs(point_1.y - point_2.y)
                           ^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'x'

Previous versions of the interpreter would point to just the line making it ambiguous which object was None. These enhanced errors can also be helpful when dealing with deeply nested dictionary objects and multiple function calls,

Traceback (most recent call last):
  File "query.py", line 37, in <module>
    magic_arithmetic('foo')
    ^^^^^^^^^^^^^^^^^^^^^^^
  File "query.py", line 18, in magic_arithmetic
    return add_counts(x) / 25
           ^^^^^^^^^^^^^
  File "query.py", line 24, in add_counts
    return 25 + query_user(user1) + query_user(user2)
                ^^^^^^^^^^^^^^^^^
  File "query.py", line 32, in query_user
    return 1 + query_count(db, response['a']['b']['c']['user'], retry=True)
                               ~~~~~~~~~~~~~~~~~~^^^^^
TypeError: 'NoneType' object is not subscriptable

as well as complex arithmetic expressions:

Traceback (most recent call last):
  File "calculation.py", line 54, in <module>
    result = (x / y / z) * (a / b / c)
              ~~~~~~^~~
ZeroDivisionError: division by zero

See 657 for more details. (Contributed by Pablo Galindo, Batuhan Taskaya and Ammar Askar in 43950.)

Note

This feature requires storing column positions in code objects which may result in a small increase of disk usage of compiled Python files or interpreter memory usage. To avoid storing the extra information and/or deactivate printing the extra traceback information, the -X no_debug_ranges command line flag or the PYTHONNODEBUGRANGES environment variable can be used.

Column information for code objects

The information used by the enhanced traceback feature is made available as a general API that can be used to correlate bytecode instructions with source code. This information can be retrieved using:

  • The codeobject.co_positions method in Python.
  • The :cPyCode_Addr2Location function in the C-API.

The -X no_debug_ranges option and the environment variable PYTHONNODEBUGRANGES can be used to disable this feature.

See 657 for more details. (Contributed by Pablo Galindo, Batuhan Taskaya and Ammar Askar in 43950.)

Exceptions can be enriched with notes (PEP 678)

The add_note method was added to BaseException. It can be used to enrich exceptions with context information which is not available at the time when the exception is raised. The notes added appear in the default traceback. See 678 for more details. (Contributed by Irit Katriel in 45607.)

New Features Related to Type Hints

This section covers major changes affecting 484 type hints and the typing module.

PEP 646: Variadic generics

484 introduced ~typing.TypeVar, enabling creation of generics parameterised with a single type. 646 introduces ~typing.TypeVarTuple, enabling parameterisation with an arbitrary number of types. In other words, a ~typing.TypeVarTuple is a variadic type variable, enabling variadic generics. This enables a wide variety of use cases. In particular, it allows the type of array-like structures in numerical computing libraries such as NumPy and TensorFlow to be parameterised with the array shape. Static type checkers will now be able to catch shape-related bugs in code that uses these libraries.

See 646 for more details.

(Contributed by Matthew Rahtz in 43224, with contributions by Serhiy Storchaka and Jelle Zijlstra. PEP written by Mark Mendoza, Matthew Rahtz, Pradeep Kumar Srinivasan, and Vincent Siles.)

PEP 655: Marking individual TypedDict items as required or not-required

~typing.Required and ~typing.NotRequired provide a straightforward way to mark whether individual items in a ~typing.TypedDict must be present. Previously this was only possible using inheritance.

Fields are still required by default, unless the total=False parameter is set. For example, the following specifies a dictionary with one required and one not-required key:

class Movie(TypedDict):
   title: str
   year: NotRequired[int]

m1: Movie = {"title": "Black Panther", "year": 2018}  # ok
m2: Movie = {"title": "Star Wars"}  # ok (year is not required)
m3: Movie = {"year": 2022}  # error (missing required field title)

The following definition is equivalent:

class Movie(TypedDict, total=False):
   title: Required[str]
   year: int

See 655 for more details.

(Contributed by David Foster and Jelle Zijlstra in 47087. PEP written by David Foster.)

PEP 673: Self type

The new ~typing.Self annotation provides a simple and intuitive way to annotate methods that return an instance of their class. This behaves the same as the ~typing.TypeVar-based approach specified in 484 but is more concise and easier to follow.

Common use cases include alternative constructors provided as classmethods and ~object.__enter__ methods that return self:

class MyLock:
    def __enter__(self) -> Self:
        self.lock()
        return self

    ...

class MyInt:
    @classmethod
    def fromhex(cls, s: str) -> Self:
        return cls(int(s, 16))

    ...

~typing.Self can also be used to annotate method parameters or attributes of the same type as their enclosing class.

See 673 for more details.

(Contributed by James Hilton-Balfe in 46534. PEP written by Pradeep Kumar Srinivasan and James Hilton-Balfe.)

PEP 675: Arbitrary literal string type

The new ~typing.LiteralString annotation may be used to indicate that a function parameter can be of any literal string type. This allows a function to accept arbitrary literal string types, as well as strings created from other literal strings. Type checkers can then enforce that sensitive functions, such as those that execute SQL statements or shell commands, are called only with static arguments, providing protection against injection attacks.

For example, a SQL query function could be annotated as follows:

def run_query(sql: LiteralString) -> ...
    ...

def caller(
    arbitrary_string: str,
    query_string: LiteralString,
    table_name: LiteralString,
) -> None:
    run_query("SELECT * FROM students")       # ok
    run_query(query_string)                   # ok
    run_query("SELECT * FROM " + table_name)  # ok
    run_query(arbitrary_string)               # type checker error
    run_query(                                # type checker error
        f"SELECT * FROM students WHERE name = {arbitrary_string}"
    )

See 675 for more details.

(Contributed by Jelle Zijlstra in 47088. PEP written by Pradeep Kumar Srinivasan and Graham Bleaney.)

PEP 681: Data Class Transforms

The new ~typing.dataclass_transform annotation may be used to decorate a function that is itself a decorator, a class, or a metaclass. The presence of @dataclass_transform() tells a static type checker that the decorated function, class, or metaclass performs runtime "magic" that transforms a class, endowing it with dataclass-like behaviors.

For example:

# The ``create_model`` decorator is defined by a library.
@typing.dataclass_transform()
def create_model(cls: Type[_T]) -> Type[_T]:
    cls.__init__ = ...
    cls.__eq__ = ...
    cls.__ne__ = ...
    return cls

# The ``create_model`` decorator can now be used to create new model
# classes, like this:
@create_model
class CustomerModel:
    id: int
    name: str

c = CustomerModel(id=327, name="John Smith")

See 681 for more details.

(Contributed by Jelle Zijlstra in 91860. PEP written by Erik De Bonte and Eric Traut.)

Other Language Changes

  • Starred expressions can be used in for statements<for>. (See 46725 for more details.)
  • Asynchronous comprehensions are now allowed inside comprehensions in asynchronous functions. Outer comprehensions implicitly become asynchronous. (Contributed by Serhiy Storchaka in 33346.)
  • A TypeError is now raised instead of an AttributeError in contextlib.ExitStack.enter_context and contextlib.AsyncExitStack.enter_async_context for objects which do not support the context manager or asynchronous context manager protocols correspondingly. (Contributed by Serhiy Storchaka in 44471.)
  • A TypeError is now raised instead of an AttributeError in with and async with statements for objects which do not support the context manager or asynchronous context manager protocols correspondingly. (Contributed by Serhiy Storchaka in 12022.)
  • Added object.__getstate__ which provides the default implementation of the __getstate__() method. Copying <copy> and pickling <pickle> instances of subclasses of builtin types bytearray, set, frozenset, collections.OrderedDict, collections.deque, weakref.WeakSet, and datetime.tzinfo now copies and pickles instance attributes implemented as slots <__slots__>. (Contributed by Serhiy Storchaka in 26579.)

Other CPython Implementation Changes

  • Special methods complex.__complex__ and bytes.__bytes__ are implemented to support typing.SupportsComplex and typing.SupportsBytes protocols. (Contributed by Mark Dickinson and Dong-hee Na in 24234.)
  • siphash13 is added as a new internal hashing algorithms. It has similar security properties as siphash24 but it is slightly faster for long inputs. str, bytes, and some other types now use it as default algorithm for hash. 552 hash-based pyc files now use siphash13, too. (Contributed by Inada Naoki in 29410.)
  • When an active exception is re-raised by a raise statement with no parameters, the traceback attached to this exception is now always sys.exc_info()[1].__traceback__. This means that changes made to the traceback in the current except clause are reflected in the re-raised exception. (Contributed by Irit Katriel in 45711.)
  • The interpreter state's representation of handled exceptions (a.k.a exc_info, or _PyErr_StackItem) now has only the exc_value field, exc_type and exc_traceback have been removed as their values can be derived from exc_value. (Contributed by Irit Katriel in 45711.)
  • A new command line option for the Windows installer AppendPath has been added. It behaves similiar to PrependPath but appends the install and scripts directories instead of prepending them. (Contributed by Bastian Neuburger in 44934.)

New Modules

  • A new module, tomllib, was added for parsing TOML. (Contributed by Taneli Hukkinen in 40059.)
  • wsgiref.types, containing WSGI-specific types for static type checking, was added. (Contributed by Sebastian Rittau in 42012.)

Improved Modules

asyncio

  • Add raw datagram socket functions to the event loop: ~asyncio.AbstractEventLoop.sock_sendto, ~asyncio.AbstractEventLoop.sock_recvfrom and ~asyncio.AbstractEventLoop.sock_recvfrom_into. (Contributed by Alex Grönholm in 46805.)
  • Add ~asyncio.streams.StreamWriter.start_tls method for upgrading existing stream-based connections to TLS. (Contributed by Ian Good in 34975.)

fractions

  • Support PEP 515-style initialization of ~fractions.Fraction from string. (Contributed by Sergey B Kirpichev in 44258.)
  • ~fractions.Fraction now implements an __int__ method, so that an isinstance(some_fraction, typing.SupportsInt) check passes. (Contributed by Mark Dickinson in 44547.)

functools

  • functools.singledispatch now supports types.UnionType and typing.Union as annotations to the dispatch argument.:

    >>> from functools import singledispatch
>>> @singledispatch
... def fun(arg, verbose=False):
...     if verbose:
...         print("Let me just say,", end=" ")
...     print(arg)
...
>>> @fun.register
... def _(arg: int | float, verbose=False):
...     if verbose:
...         print("Strength in numbers, eh?", end=" ")
...     print(arg)
...
>>> from typing import Union
>>> @fun.register
... def _(arg: Union[list, set], verbose=False):
...     if verbose:
...         print("Enumerate this:")
...     for i, elem in enumerate(arg):
...         print(i, elem)
...

    (Contributed by Yurii Karabas in 46014.)

hashlib

  • hashlib.blake2b and hashlib.blake2s now prefer libb2 over Python's vendored copy. (Contributed by Christian Heimes in 47095.)
  • The internal _sha3 module with SHA3 and SHAKE algorithms now uses tiny_sha3 instead of the Keccak Code Package to reduce code and binary size. The hashlib module prefers optimized SHA3 and SHAKE implementations from OpenSSL. The change affects only installations without OpenSSL support. (Contributed by Christian Heimes in 47098.)

IDLE and idlelib

  • Apply syntax highlighting to .pyi files. (Contributed by Alex Waygood and Terry Jan Reedy in 45447.)

inspect

  • Add inspect.getmembers_static: return all members without triggering dynamic lookup via the descriptor protocol. (Contributed by Weipeng Hong in 30533.)
  • Add inspect.ismethodwrapper for checking if the type of an object is a ~types.MethodWrapperType. (Contributed by Hakan Çelik in 29418.)
  • Change the frame-related functions in the inspect module to return a regular object (that is backwards compatible with the old tuple-like interface) that include the extended 657 position information (end line number, column and end column). The affected functions are: inspect.getframeinfo, inspect.getouterframes, inspect.getinnerframes, inspect.stack and inspect.trace. (Contributed by Pablo Galindo in 88116)

locale

  • Add locale.getencoding to get the current locale encoding. It is similar to locale.getpreferredencoding(False) but ignores the Python UTF-8 Mode <utf8-mode>.

math

  • Add math.exp2: return 2 raised to the power of x. (Contributed by Gideon Mitchell in 45917.)
  • Add math.cbrt: return the cube root of x. (Contributed by Ajith Ramachandran in 44357.)
  • The behaviour of two math.pow corner cases was changed, for consistency with the IEEE 754 specification. The operations math.pow(0.0, -math.inf) and math.pow(-0.0, -math.inf) now return inf. Previously they raised ValueError. (Contributed by Mark Dickinson in 44339.)
  • The math.nan value is now always available. (Contributed by Victor Stinner in 46917.)

operator

  • A new function operator.call has been added, such that operator.call(obj, *args, **kwargs) == obj(*args, **kwargs). (Contributed by Antony Lee in 44019.)

os

  • On Windows, os.urandom now uses BCryptGenRandom(), instead of CryptGenRandom() which is deprecated. (Contributed by Dong-hee Na in 44611.)

pathlib

  • ~pathlib.Path.glob and ~pathlib.Path.rglob return only directories if pattern ends with a pathname components separator: ~os.sep or ~os.altsep. (Contributed by Eisuke Kawasima in 22276 and 33392.)

re

  • Atomic grouping ((?>...)) and possessive quantifiers (*+, ++, ?+, {m,n}+) are now supported in regular expressions. (Contributed by Jeffrey C. Jacobs and Serhiy Storchaka in 433030.)

shutil

  • Add optional parameter dir_fd in shutil.rmtree. (Contributed by Serhiy Storchaka in 46245.)

socket

  • Add CAN Socket support for NetBSD. (Contributed by Thomas Klausner in 30512.)
  • ~socket.create_connection has an option to raise, in case of failure to connect, an ExceptionGroup containing all errors instead of only raising the last error. (Contributed by Irit Katriel in 29980).

sqlite3

  • You can now disable the authorizer by passing None to ~sqlite3.Connection.set_authorizer. (Contributed by Erlend E. Aasland in 44491.)
  • Collation name ~sqlite3.Connection.create_collation can now contain any Unicode character. Collation names with invalid characters now raise UnicodeEncodeError instead of sqlite3.ProgrammingError. (Contributed by Erlend E. Aasland in 44688.)
  • sqlite3 exceptions now include the SQLite extended error code as ~sqlite3.Error.sqlite_errorcode and the SQLite error name as ~sqlite3.Error.sqlite_errorname. (Contributed by Aviv Palivoda, Daniel Shahaf, and Erlend E. Aasland in 16379 and 24139.)
  • Add ~sqlite3.Connection.setlimit and ~sqlite3.Connection.getlimit to sqlite3.Connection for setting and getting SQLite limits by connection basis. (Contributed by Erlend E. Aasland in 45243.)
  • sqlite3 now sets sqlite3.threadsafety based on the default threading mode the underlying SQLite library has been compiled with. (Contributed by Erlend E. Aasland in 45613.)
  • sqlite3 C callbacks now use unraisable exceptions if callback tracebacks are enabled. Users can now register an unraisable hook handler <sys.unraisablehook> to improve their debug experience. (Contributed by Erlend E. Aasland in 45828.)
  • Fetch across rollback no longer raises ~sqlite3.InterfaceError. Instead we leave it to the SQLite library to handle these cases. (Contributed by Erlend E. Aasland in 44092.)
  • Add ~sqlite3.Connection.serialize and ~sqlite3.Connection.deserialize to sqlite3.Connection for serializing and deserializing databases. (Contributed by Erlend E. Aasland in 41930.)
  • Add ~sqlite3.Connection.create_window_function to sqlite3.Connection for creating aggregate window functions. (Contributed by Erlend E. Aasland in 34916.)
  • Add ~sqlite3.Connection.blobopen to sqlite3.Connection. sqlite3.Blob allows incremental I/O operations on blobs. (Contributed by Aviv Palivoda and Erlend E. Aasland in 24905)

sys

  • sys.exc_info now derives the type and traceback fields from the value (the exception instance), so when an exception is modified while it is being handled, the changes are reflected in the results of subsequent calls to exc_info. (Contributed by Irit Katriel in 45711.)
  • Add sys.exception which returns the active exception instance (equivalent to sys.exc_info()[1]). (Contributed by Irit Katriel in 46328.)

sysconfig

  • Two new installation schemes <installation_paths> (posix_venv, nt_venv and venv) were added and are used when Python creates new virtual environments or when it is running from a virtual environment. The first two schemes (posix_venv and nt_venv) are OS-specific for non-Windows and Windows, the venv is essentially an alias to one of them according to the OS Python runs on. This is useful for downstream distributors who modify sysconfig.get_preferred_scheme. Third party code that creates new virtual environments should use the new venv installation scheme to determine the paths, as does venv. (Contributed by Miro Hrončok in 45413.)

threading

  • On Unix, if the sem_clockwait() function is available in the C library (glibc 2.30 and newer), the threading.Lock.acquire method now uses the monotonic clock (time.CLOCK_MONOTONIC) for the timeout, rather than using the system clock (time.CLOCK_REALTIME), to not be affected by system clock changes. (Contributed by Victor Stinner in 41710.)

time

  • On Unix, time.sleep now uses the clock_nanosleep() or nanosleep() function, if available, which has a resolution of 1 nanosecond (10-9 seconds), rather than using select() which has a resolution of 1 microsecond (10-6 seconds). (Contributed by Benjamin Szőke and Victor Stinner in 21302.)
  • On Windows 8.1 and newer, time.sleep now uses a waitable timer based on high-resolution timers which has a resolution of 100 nanoseconds (10-7 seconds). Previously, it had a resolution of 1 millisecond (10-3 seconds). (Contributed by Benjamin Szőke, Dong-hee Na, Eryk Sun and Victor Stinner in 21302 and 45429.)

typing

For major changes, see new-feat-related-type-hints-311.

  • Add typing.assert_never and typing.Never. typing.assert_never is useful for asking a type checker to confirm that a line of code is not reachable. At runtime, it raises an AssertionError. (Contributed by Jelle Zijlstra in 90633.)
  • Add typing.reveal_type. This is useful for asking a type checker what type it has inferred for a given expression. At runtime it prints the type of the received value. (Contributed by Jelle Zijlstra in 90572.)
  • Add typing.assert_type. This is useful for asking a type checker to confirm that the type it has inferred for a given expression matches the given type. At runtime it simply returns the received value. (Contributed by Jelle Zijlstra in 90638.)
  • Allow subclassing of typing.Any. This is useful for avoiding type checker errors related to highly dynamic class, such as mocks. (Contributed by Shantanu Jain in 91154.)
  • The typing.final decorator now sets the __final__ attributed on the decorated object. (Contributed by Jelle Zijlstra in 90500.)
  • The typing.get_overloads function can be used for introspecting the overloads of a function. typing.clear_overloads can be used to clear all registered overloads of a function. (Contributed by Jelle Zijlstra in 89263.)
  • ~typing.NamedTuple subclasses can be generic. (Contributed by Serhiy Storchaka in 43923.)

unicodedata

  • The Unicode database has been updated to version 14.0.0. (45190).

venv

  • When new Python virtual environments are created, the venv sysconfig installation scheme <installation_paths> is used to determine the paths inside the environment. When Python runs in a virtual environment, the same installation scheme is the default. That means that downstream distributors can change the default sysconfig install scheme without changing behavior of virtual environments. Third party code that also creates new virtual environments should do the same. (Contributed by Miro Hrončok in 45413.)

warnings

  • warnings.catch_warnings now accepts arguments for warnings.simplefilter, providing a more concise way to locally ignore warnings or convert them to errors. (Contributed by Zac Hatfield-Dodds in 47074.)

zipfile

  • Added support for specifying member name encoding for reading metadata in the zipfile's directory and file headers. (Contributed by Stephen J. Turnbull and Serhiy Storchaka in 28080.)

fcntl

  • On FreeBSD, the F_DUP2FD and F_DUP2FD_CLOEXEC flags respectively are supported, the former equals to dup2 usage while the latter set the FD_CLOEXEC flag in addition.

Optimizations

  • Compiler now optimizes simple C-style formatting with literal format containing only format codes %s, %r and %a and makes it as fast as corresponding f-string expression. (Contributed by Serhiy Storchaka in 28307.)
  • "Zero-cost" exceptions are implemented. The cost of try statements is almost eliminated when no exception is raised. (Contributed by Mark Shannon in 40222.)
  • Pure ASCII strings are now normalized in constant time by unicodedata.normalize. (Contributed by Dong-hee Na in 44987.)
  • math functions ~math.comb and ~math.perm are now up to 10 times or more faster for large arguments (the speed up is larger for larger k). (Contributed by Serhiy Storchaka in 37295.)
  • Dict don't store hash value when all inserted keys are Unicode objects. This reduces dict size. For example, sys.getsizeof(dict.fromkeys("abcdefg")) becomes 272 bytes from 352 bytes on 64bit platform. (Contributed by Inada Naoki in 46845.)
  • re's regular expression matching engine has been partially refactored, and now uses computed gotos (or "threaded code") on supported platforms. As a result, Python 3.11 executes the pyperformance regular expression benchmarks up to 10% faster than Python 3.10.

Faster CPython

CPython 3.11 is on average 1.22x faster than CPython 3.10 when measured with the pyperformance benchmark suite, and compiled with GCC on Ubuntu Linux. Depending on your workload, the speedup could be up to 10-60% faster.

This project focuses on two major areas in Python: faster startup and faster runtime. Other optimizations not under this project are listed in Optimizations.

Faster Startup

Frozen imports / Static code objects

Python caches bytecode in the __pycache__<tut-pycache> directory to speed up module loading.

Previously in 3.10, Python module execution looked like this:

Read __pycache__ -> Unmarshal -> Heap allocated code object -> Evaluate

In Python 3.11, the core modules essential for Python startup are "frozen". This means that their code objects (and bytecode) are statically allocated by the interpreter. This reduces the steps in module execution process to this:

Statically allocated code object -> Evaluate

Interpreter startup is now 10-15% faster in Python 3.11. This has a big impact for short-running programs using Python.

(Contributed by Eric Snow, Guido van Rossum and Kumar Aditya in numerous issues.)

Faster Runtime

Cheaper, lazy Python frames

Python frames are created whenever Python calls a Python function. This frame holds execution information. The following are new frame optimizations:

  • Streamlined the frame creation process.
  • Avoided memory allocation by generously re-using frame space on the C stack.
  • Streamlined the internal frame struct to contain only essential information. Frames previously held extra debugging and memory management information.

Old-style frame objects are now created only when required by debuggers. For most user code, no frame objects are created at all. As a result, nearly all Python functions calls have sped up significantly. We measured a 3-7% speedup in pyperformance.

(Contributed by Mark Shannon in 44590.)

Inlined Python function calls

During a Python function call, Python will call an evaluating C function to interpret that function's code. This effectively limits pure Python recursion to what's safe for the C stack.

In 3.11, when CPython detects Python code calling another Python function, it sets up a new frame, and "jumps" to the new code inside the new frame. This avoids calling the C interpreting function altogether.

Most Python function calls now consume no C stack space. This speeds up most of such calls. In simple recursive functions like fibonacci or factorial, a 1.7x speedup was observed. This also means recursive functions can recurse significantly deeper (if the user increases the recursion limit). We measured a 1-3% improvement in pyperformance.

(Contributed by Pablo Galindo and Mark Shannon in 45256.)

PEP 659: Specializing Adaptive Interpreter

659 is one of the key parts of the faster CPython project. The general idea is that while Python is a dynamic language, most code has regions where objects and types rarely change. This concept is known as type stability.

At runtime, Python will try to look for common patterns and type stability in the executing code. Python will then replace the current operation with a more specialized one. This specialized operation uses fast paths available only to those use cases/types, which generally outperform their generic counterparts. This also brings in another concept called inline caching, where Python caches the results of expensive operations directly in the bytecode.

The specializer will also combine certain common instruction pairs into one superinstruction. This reduces the overhead during execution.

Python will only specialize when it sees code that is "hot" (executed multiple times). This prevents Python from wasting time for run-once code. Python can also de-specialize when code is too dynamic or when the use changes. Specialization is attempted periodically, and specialization attempts are not too expensive. This allows specialization to adapt to new circumstances.

(PEP written by Mark Shannon, with ideas inspired by Stefan Brunthaler. See 659 for more information.)

Operation Form Specialization Operation speedup (up to) Contributor(s)
Binary operations x+x; x*x; x-x; Binary add, multiply and subtract for common types such as int, float, and str take custom fast paths for their underlying types. 10% Mark Shannon, Dong-hee Na, Brandt Bucher, Dennis Sweeney
Subscript a[i]

Subscripting container types such as list, tuple and dict directly index the underlying data structures.

Subscripting custom __getitem__ is also inlined similar to inline-calls.

 10-25% Irit Katriel, Mark Shannon
Store subscript a[i] = z Similar to subscripting specialization above. 10-25% Dennis Sweeney
Calls f(arg) C(arg) Calls to common builtin (C) functions and types such as len and str directly call their underlying C version. This avoids going through the internal calling convention. 20% Mark Shannon, Ken Jin
Load global variable print len The object's index in the globals/builtins namespace is cached. Loading globals and builtins require zero namespace lookups. 1 Mark Shannon
Load attribute o.attr Similar to loading global variables. The attribute's index inside the class/object's namespace is cached. In most cases, attribute loading will require zero namespace lookups. 2 Mark Shannon
Load methods for call o.meth() The actual address of the method is cached. Method loading now has no namespace lookups -- even for classes with long inheritance chains. 10-20% Ken Jin, Mark Shannon
Store attribute o.attr = z Similar to load attribute optimization. 2% in pyperformance Mark Shannon
Unpack Sequence *seq Specialized for common containers such as list and tuple. Avoids internal calling convention. 8% Brandt Bucher

Misc

  • Objects now require less memory due to lazily created object namespaces. Their namespace dictionaries now also share keys more freely. (Contributed Mark Shannon in 45340 and 40116.)
  • A more concise representation of exceptions in the interpreter reduced the time required for catching an exception by about 10%. (Contributed by Irit Katriel in 45711.)

FAQ

Q: How should I write my code to utilize these speedups?

A: You don't have to change your code. Write Pythonic code that follows common best practices. The Faster CPython project optimizes for common code patterns we observe.


Q: Will CPython 3.11 use more memory?

A: Maybe not. We don't expect memory use to exceed 20% more than 3.10. This is offset by memory optimizations for frame objects and object dictionaries as mentioned above.


Q: I don't see any speedups in my workload. Why?

A: Certain code won't have noticeable benefits. If your code spends most of its time on I/O operations, or already does most of its computation in a C extension library like numpy, there won't be significant speedup. This project currently benefits pure-Python workloads the most.

Furthermore, the pyperformance figures are a geometric mean. Even within the pyperformance benchmarks, certain benchmarks have slowed down slightly, while others have sped up by nearly 2x!


Q: Is there a JIT compiler?

A: No. We're still exploring other optimizations.

About

Faster CPython explores optimizations for CPython. The main team is funded by Microsoft to work on this full-time. Pablo Galindo Salgado is also funded by Bloomberg LP to work on the project part-time. Finally, many contributors are volunteers from the community.

CPython bytecode changes

  • Replaced all numeric BINARY_* and INPLACE_* instructions with a single BINARY_OP implementation.
  • Replaced the three call instructions: CALL_FUNCTION, CALL_FUNCTION_KW and CALL_METHOD with PUSH_NULL, PRECALL, CALL, and KW_NAMES. This decouples the argument shifting for methods from the handling of keyword arguments and allows better specialization of calls.
  • Removed COPY_DICT_WITHOUT_KEYS and GEN_START.
  • MATCH_CLASS and MATCH_KEYS no longer push an additional boolean value indicating whether the match succeeded or failed. Instead, they indicate failure with None (where a tuple of extracted values would otherwise be).
  • Replace several stack manipulation instructions (DUP_TOP, DUP_TOP_TWO, ROT_TWO, ROT_THREE, ROT_FOUR, and ROT_N) with new COPY and SWAP instructions.
  • Replaced JUMP_IF_NOT_EXC_MATCH by CHECK_EXC_MATCH which performs the check but does not jump.
  • Replaced JUMP_IF_NOT_EG_MATCH by CHECK_EG_MATCH which performs the check but does not jump.
  • Replaced JUMP_ABSOLUTE by the relative JUMP_BACKWARD.
  • Added JUMP_BACKWARD_NO_INTERRUPT, which is used in certain loops where it is undesirable to handle interrupts.
  • Replaced POP_JUMP_IF_TRUE and POP_JUMP_IF_FALSE by the relative POP_JUMP_FORWARD_IF_TRUE, POP_JUMP_BACKWARD_IF_TRUE, POP_JUMP_FORWARD_IF_FALSE and POP_JUMP_BACKWARD_IF_FALSE.
  • Added POP_JUMP_FORWARD_IF_NOT_NONE, POP_JUMP_BACKWARD_IF_NOT_NONE, POP_JUMP_FORWARD_IF_NONE and POP_JUMP_BACKWARD_IF_NONE opcodes to speed up conditional jumps.
  • JUMP_IF_TRUE_OR_POP and JUMP_IF_FALSE_OR_POP are now relative rather than absolute.

Deprecated

  • Octal escapes with value larger than 0o377 now produce a DeprecationWarning. In a future Python version they will be a SyntaxWarning and eventually a SyntaxError. (Contributed by Serhiy Storchaka in 81548.)
  • The lib2to3 package and 2to3 tool are now deprecated and may not be able to parse Python 3.10 or newer. See the 617 (New PEG parser for CPython). (Contributed by Victor Stinner in 40360.)
  • Undocumented modules sre_compile, sre_constants and sre_parse are now deprecated. (Contributed by Serhiy Storchaka in 47152.)
  • webbrowser.MacOSX is deprecated and will be removed in Python 3.13. It is untested and undocumented and also not used by webbrowser itself. (Contributed by Dong-hee Na in 42255.)
  • The behavior of returning a value from a ~unittest.TestCase and ~unittest.IsolatedAsyncioTestCase test methods (other than the default None value), is now deprecated.

  • Deprecated the following unittest functions, scheduled for removal in Python 3.13:

    • unittest.findTestCases
    • unittest.makeSuite
    • unittest.getTestCaseNames

    Use ~unittest.TestLoader method instead:

    • unittest.TestLoader.loadTestsFromModule
    • unittest.TestLoader.loadTestsFromTestCase
    • unittest.TestLoader.getTestCaseNames

    (Contributed by Erlend E. Aasland in 5846.)

  • The turtle.RawTurtle.settiltangle is deprecated since Python 3.1, it now emits a deprecation warning and will be removed in Python 3.13. Use turtle.RawTurtle.tiltangle instead (it was earlier incorrectly marked as deprecated, its docstring is now corrected). (Contributed by Hugo van Kemenade in 45837.)
  • The delegation of int to __trunc__ is now deprecated. Calling int(a) when type(a) implements __trunc__ but not __int__ or __index__ now raises a DeprecationWarning. (Contributed by Zackery Spytz in 44977.)

  • The following have been deprecated in configparser since Python 3.2. Their deprecation warnings have now been updated to note they will removed in Python 3.12:

    • the configparser.SafeConfigParser class
    • the configparser.ParsingError.filename property
    • the configparser.ParsingError.readfp method

    (Contributed by Hugo van Kemenade in 45173.)

  • configparser.LegacyInterpolation has been deprecated in the docstring since Python 3.2. It now emits a DeprecationWarning and will be removed in Python 3.13. Use configparser.BasicInterpolation or configparser.ExtendedInterpolation instead. (Contributed by Hugo van Kemenade in 46607.)
  • The locale.getdefaultlocale function is deprecated and will be removed in Python 3.13. Use locale.setlocale, locale.getpreferredencoding(False) <locale.getpreferredencoding> and locale.getlocale functions instead. (Contributed by Victor Stinner in 46659.)
  • The asynchat, asyncore and smtpd modules have been deprecated since at least Python 3.6. Their documentation and deprecation warnings have now been updated to note they will removed in Python 3.12 (594). (Contributed by Hugo van Kemenade in 47022.)

  • 594 led to the deprecations of the following modules which are slated for removal in Python 3.13:

    • aifc
    • audioop
    • cgi
    • cgitb
    • chunk
    • crypt
    • imghdr
    • mailcap
    • msilib
    • nis
    • nntplib
    • ossaudiodev
    • pipes
    • sndhdr
    • spwd
    • sunau
    • telnetlib
    • uu
    • xdrlib

    (Contributed by Brett Cannon in 47061 and Victor Stinner in 68966.)

  • More strict rules will be applied now applied for numerical group references and group names in regular expressions in future Python versions. Only sequence of ASCII digits will be now accepted as a numerical reference. The group name in bytes patterns and replacement strings could only contain ASCII letters and digits and underscore. For now, a deprecation warning is raised for such syntax. (Contributed by Serhiy Storchaka in 91760.)

Removed

  • smtpd.MailmanProxy is now removed as it is unusable without an external module, mailman. (Contributed by Dong-hee Na in 35800.)

  • The binhex module, deprecated in Python 3.9, is now removed. The following binascii functions, deprecated in Python 3.9, are now also removed:

    • a2b_hqx(), b2a_hqx();
    • rlecode_hqx(), rledecode_hqx().

    The binascii.crc_hqx function remains available.

    (Contributed by Victor Stinner in 45085.)

  • The distutils bdist_msi command, deprecated in Python 3.9, is now removed. Use bdist_wheel (wheel packages) instead. (Contributed by Hugo van Kemenade in 45124.)
  • Due to significant security concerns, the reuse_address parameter of asyncio.loop.create_datagram_endpoint, disabled in Python 3.9, is now entirely removed. This is because of the behavior of the socket option SO_REUSEADDR in UDP. (Contributed by Hugo van Kemenade in 45129.)
  • Removed __getitem__ methods of xml.dom.pulldom.DOMEventStream, wsgiref.util.FileWrapper and fileinput.FileInput, deprecated since Python 3.9. (Contributed by Hugo van Kemenade in 45132.)

  • The following deprecated functions and methods are removed in the gettext module: ~gettext.lgettext, ~gettext.ldgettext, ~gettext.lngettext and ~gettext.ldngettext.

    Function ~gettext.bind_textdomain_codeset, methods ~gettext.NullTranslations.output_charset and ~gettext.NullTranslations.set_output_charset, and the codeset parameter of functions ~gettext.translation and ~gettext.install are also removed, since they are only used for the l*gettext() functions. (Contributed by Dong-hee Na and Serhiy Storchaka in 44235.)

  • The @asyncio.coroutine <asyncio.coroutine> decorator enabling legacy generator-based coroutines to be compatible with async/await code. The function has been deprecated since Python 3.8 and the removal was initially scheduled for Python 3.10. Use async def instead. (Contributed by Illia Volochii in 43216.)
  • asyncio.coroutines.CoroWrapper used for wrapping legacy generator-based coroutine objects in the debug mode. (Contributed by Illia Volochii in 43216.)
  • Removed the deprecated split() method of _tkinter.TkappType. (Contributed by Erlend E. Aasland in 38371.)

  • Removed from the inspect module:

    • the getargspec function, deprecated since Python 3.0; use inspect.signature or inspect.getfullargspec instead.
    • the formatargspec function, deprecated since Python 3.5; use the inspect.signature function and Signature object directly.
    • the undocumented Signature.from_builtin and Signature.from_function functions, deprecated since Python 3.5; use the Signature.from_callable() <inspect.Signature.from_callable> method instead.

    (Contributed by Hugo van Kemenade in 45320.)

  • Remove namespace package support from unittest discovery. It was introduced in Python 3.4 but has been broken since Python 3.7. (Contributed by Inada Naoki in 23882.)
  • Remove __class_getitem__ method from pathlib.PurePath, because it was not used and added by mistake in previous versions. (Contributed by Nikita Sobolev in 46483.)
  • Remove the undocumented private float.__set_format__() method, previously known as float.__setformat__() in Python 3.7. Its docstring said: "You probably don't want to use this function. It exists mainly to be used in Python's test suite." (Contributed by Victor Stinner in 46852.)

Porting to Python 3.11

This section lists previously described changes and other bugfixes that may require changes to your code.

Changes in the Python API

  • Prohibited passing non-concurrent.futures.ThreadPoolExecutor executors to loop.set_default_executor following a deprecation in Python 3.8. (Contributed by Illia Volochii in 43234.)
  • open, io.open, codecs.open and fileinput.FileInput no longer accept 'U' ("universal newline") in the file mode. This flag was deprecated since Python 3.3. In Python 3, the "universal newline" is used by default when a file is open in text mode. The newline parameter <open-newline-parameter> of open controls how universal newlines works. (Contributed by Victor Stinner in 37330.)
  • The pdb module now reads the .pdbrc configuration file with the 'utf-8' encoding. (Contributed by Srinivas Reddy Thatiparthy (శ్రీనివాస్ రెడ్డి తాటిపర్తి) in 41137.)
  • When sorting using tuples as keys, the order of the result may differ from earlier releases if the tuple elements don't define a total ordering (see expressions-value-comparisons for information on total ordering). It's generally true that the result of sorting simply isn't well-defined in the absence of a total ordering on list elements.
  • calendar: The calendar.LocaleTextCalendar and calendar.LocaleHTMLCalendar classes now use locale.getlocale, instead of using locale.getdefaultlocale, if no locale is specified. (Contributed by Victor Stinner in 46659.)
  • Global inline flags (e.g. (?i)) can now only be used at the start of the regular expressions. Using them not at the start of expression was deprecated since Python 3.6. (Contributed by Serhiy Storchaka in 47066.)
  • re module: Fix a few long-standing bugs where, in rare cases, capturing group could get wrong result. So the result may be different than before. (Contributed by Ma Lin in 35859.)
  • The population parameter of random.sample must be a sequence. Automatic conversion of sets to lists is no longer supported. If the sample size is larger than the population size, a ValueError is raised. (Contributed by Raymond Hettinger in 40465.)

Build Changes

  • Building Python now requires a C11 compiler without optional C11 features. (Contributed by Victor Stinner in 46656.)
  • Building Python now requires support of IEEE 754 floating point numbers. (Contributed by Victor Stinner in 46917.)
  • CPython can now be built with the ThinLTO option via --with-lto=thin. (Contributed by Dong-hee Na and Brett Holman in 44340.)
  • libpython is no longer linked against libcrypt. (Contributed by Mike Gilbert in 45433.)
  • Building Python now requires a C99 <math.h> header file providing the following functions: copysign(), hypot(), isfinite(), isinf(), isnan(), round(). (Contributed by Victor Stinner in 45440.)
  • Building Python now requires a C99 <math.h> header file providing a NAN constant, or the __builtin_nan() built-in function. (Contributed by Victor Stinner in 46640.)
  • Building Python now requires support for floating point Not-a-Number (NaN): remove the Py_NO_NAN macro. (Contributed by Victor Stinner in 46656.)
  • Freelists for object structs can now be disabled. A new configure option !--without-freelists can be used to disable all freelists except empty tuple singleton. (Contributed by Christian Heimes in 45522)
  • Modules/Setup and Modules/makesetup have been improved and tied up. Extension modules can now be built through makesetup. All except some test modules can be linked statically into main binary or library. (Contributed by Brett Cannon and Christian Heimes in 45548, 45570, 45571, and 43974.)

  • Build dependencies, compiler flags, and linker flags for most stdlib extension modules are now detected by configure. libffi, libnsl, libsqlite3, zlib, bzip2, liblzma, libcrypt, Tcl/Tk libs, and uuid flags are detected by pkg-config (when available). (Contributed by Christian Heimes and Erlend Egeberg Aasland in 45847, 45747, and 45763.)

    Note

    Use the environment variables TCLTK_CFLAGS and TCLTK_LIBS to manually specify the location of Tcl/Tk headers and libraries. The configure options --with-tcltk-includes and --with-tcltk-libs have been removed.

  • CPython now has experimental support for cross compiling to WebAssembly platform wasm32-emscripten. The effort is inspired by previous work like Pyodide. (Contributed by Christian Heimes and Ethan Smith in 40280.)
  • CPython will now use 30-bit digits by default for the Python int implementation. Previously, the default was to use 30-bit digits on platforms with SIZEOF_VOID_P >= 8, and 15-bit digits otherwise. It's still possible to explicitly request use of 15-bit digits via either the --enable-big-digits option to the configure script or (for Windows) the PYLONG_BITS_IN_DIGIT variable in PC/pyconfig.h, but this option may be removed at some point in the future. (Contributed by Mark Dickinson in 45569.)
  • The tkinter package now requires Tcl/Tk version 8.5.12 or newer. (Contributed by Serhiy Storchaka in 46996.)

C API Changes

  • :cPyErr_SetExcInfo() no longer uses the type and traceback arguments, the interpreter now derives those values from the exception instance (the value argument). The function still steals references of all three arguments. (Contributed by Irit Katriel in 45711.)
  • :cPyErr_GetExcInfo() now derives the type and traceback fields of the result from the exception instance (the value field). (Contributed by Irit Katriel in 45711.)
  • :c_frozen has a new is_package field to indicate whether or not the frozen module is a package. Previously, a negative value in the size field was the indicator. Now only non-negative values be used for size. (Contributed by Kumar Aditya in 46608.)
  • :c_PyFrameEvalFunction now takes _PyInterpreterFrame* as its second parameter, instead of PyFrameObject*. See 523 for more details of how to use this function pointer type.
  • :cPyCode_New and :cPyCode_NewWithPosOnlyArgs now take an additional exception_table argument. Using these functions should be avoided, if at all possible. To get a custom code object: create a code object using the compiler, then get a modified version with the replace method.
  • :cPyCodeObject no longer has a co_code field. Instead, use PyObject_GetAttrString(code_object, "co_code") or :cPyCode_GetCode to get the underlying bytes object. (Contributed by Brandt Bucher in 46841 and Ken Jin in 92154.)

New Features

  • Add a new :cPyType_GetName function to get type's short name. (Contributed by Hai Shi in 42035.)
  • Add a new :cPyType_GetQualName function to get type's qualified name. (Contributed by Hai Shi in 42035.)
  • Add new :cPyThreadState_EnterTracing and :cPyThreadState_LeaveTracing functions to the limited C API to suspend and resume tracing and profiling. (Contributed by Victor Stinner in 43760.)
  • Added the :cPy_Version constant which bears the same value as :cPY_VERSION_HEX. (Contributed by Gabriele N. Tornetta in 43931.)

  • :cPy_buffer and APIs are now part of the limited API and the stable ABI:

    • :cPyObject_CheckBuffer
    • :cPyObject_GetBuffer
    • :cPyBuffer_GetPointer
    • :cPyBuffer_SizeFromFormat
    • :cPyBuffer_ToContiguous
    • :cPyBuffer_FromContiguous
    • :cPyBuffer_CopyData
    • :cPyBuffer_IsContiguous
    • :cPyBuffer_FillContiguousStrides
    • :cPyBuffer_FillInfo
    • :cPyBuffer_Release
    • :cPyMemoryView_FromBuffer
    • :c~PyBufferProcs.bf_getbuffer and :c~PyBufferProcs.bf_releasebuffer type slots

    (Contributed by Christian Heimes in 45459.)

  • Added the :cPyType_GetModuleByDef function, used to get the module in which a method was defined, in cases where this information is not available directly (via :cPyCMethod). (Contributed by Petr Viktorin in 46613.)
  • Add new functions to pack and unpack C double (serialize and deserialize): :cPyFloat_Pack2, :cPyFloat_Pack4, :cPyFloat_Pack8, :cPyFloat_Unpack2, :cPyFloat_Unpack4 and :cPyFloat_Unpack8. (Contributed by Victor Stinner in 46906.)
  • Add new functions to get frame object attributes: :cPyFrame_GetBuiltins, :cPyFrame_GetGenerator, :cPyFrame_GetGlobals, :cPyFrame_GetLasti.
  • Added two new functions to get and set the active exception instance: :cPyErr_GetHandledException and :cPyErr_SetHandledException. These are alternatives to :cPyErr_SetExcInfo() and :cPyErr_GetExcInfo() which work with the legacy 3-tuple representation of exceptions. (Contributed by Irit Katriel in 46343.)

Porting to Python 3.11

  • The old trashcan macros (Py_TRASHCAN_SAFE_BEGIN/Py_TRASHCAN_SAFE_END) are now deprecated. They should be replaced by the new macros Py_TRASHCAN_BEGIN and Py_TRASHCAN_END.

    A tp_dealloc function that has the old macros, such as:

    static void
mytype_dealloc(mytype *p)
{
    PyObject_GC_UnTrack(p);
    Py_TRASHCAN_SAFE_BEGIN(p);
    ...
    Py_TRASHCAN_SAFE_END
}

    should migrate to the new macros as follows:

    static void
mytype_dealloc(mytype *p)
{
    PyObject_GC_UnTrack(p);
    Py_TRASHCAN_BEGIN(p, mytype_dealloc)
    ...
    Py_TRASHCAN_END
}

    Note that Py_TRASHCAN_BEGIN has a second argument which should be the deallocation function it is in.

    To support older Python versions in the same codebase, you can define the following macros and use them throughout the code (credit: these were copied from the mypy codebase):

    #if PY_MAJOR_VERSION >= 3 && PY_MINOR_VERSION >= 8
#  define CPy_TRASHCAN_BEGIN(op, dealloc) Py_TRASHCAN_BEGIN(op, dealloc)
#  define CPy_TRASHCAN_END(op) Py_TRASHCAN_END
#else
#  define CPy_TRASHCAN_BEGIN(op, dealloc) Py_TRASHCAN_SAFE_BEGIN(op)
#  define CPy_TRASHCAN_END(op) Py_TRASHCAN_SAFE_END(op)
#endif
  • The :cPyType_Ready function now raises an error if a type is defined with the Py_TPFLAGS_HAVE_GC flag set but has no traverse function (:cPyTypeObject.tp_traverse). (Contributed by Victor Stinner in 44263.)
  • Heap types with the Py_TPFLAGS_IMMUTABLETYPE flag can now inherit the 590 vectorcall protocol. Previously, this was only possible for static types <static-types>. (Contributed by Erlend E. Aasland in 43908)

  • Since :cPy_TYPE() is changed to a inline static function, Py_TYPE(obj) = new_type must be replaced with Py_SET_TYPE(obj, new_type): see the :cPy_SET_TYPE() function (available since Python 3.9). For backward compatibility, this macro can be used:

    #if PY_VERSION_HEX < 0x030900A4 && !defined(Py_SET_TYPE)
static inline void _Py_SET_TYPE(PyObject *ob, PyTypeObject *type)
{ ob->ob_type = type; }
#define Py_SET_TYPE(ob, type) _Py_SET_TYPE((PyObject*)(ob), type)
#endif

    (Contributed by Victor Stinner in 39573.)

  • Since :cPy_SIZE() is changed to a inline static function, Py_SIZE(obj) = new_size must be replaced with Py_SET_SIZE(obj, new_size): see the :cPy_SET_SIZE() function (available since Python 3.9). For backward compatibility, this macro can be used:

    #if PY_VERSION_HEX < 0x030900A4 && !defined(Py_SET_SIZE)
static inline void _Py_SET_SIZE(PyVarObject *ob, Py_ssize_t size)
{ ob->ob_size = size; }
#define Py_SET_SIZE(ob, size) _Py_SET_SIZE((PyVarObject*)(ob), size)
#endif

    (Contributed by Victor Stinner in 39573.)

  • <Python.h> no longer includes the header files <stdlib.h>, <stdio.h>, <errno.h> and <string.h> when the Py_LIMITED_API macro is set to 0x030b0000 (Python 3.11) or higher. C extensions should explicitly include the header files after #include <Python.h>. (Contributed by Victor Stinner in 45434.)
  • The non-limited API files cellobject.h, classobject.h, code.h, context.h, funcobject.h, genobject.h and longintrepr.h have been moved to the Include/cpython directory. Moreover, the eval.h header file was removed. These files must not be included directly, as they are already included in Python.h: Include Files <api-includes>. If they have been included directly, consider including Python.h instead. (Contributed by Victor Stinner in 35134.)
  • The :cPyUnicode_CHECK_INTERNED macro has been excluded from the limited C API. It was never usable there, because it used internal structures which are not available in the limited C API. (Contributed by Victor Stinner in 46007.)

  • The :cPyFrameObject structure members have been removed from the public C API.

    While the documentation notes that the :cPyFrameObject fields are subject to change at any time, they have been stable for a long time and were used in several popular extensions.

    In Python 3.11, the frame struct was reorganized to allow performance optimizations. Some fields were removed entirely, as they were details of the old implementation.

    :cPyFrameObject fields:

    • f_back: use :cPyFrame_GetBack.
    • f_blockstack: removed.
    • f_builtins: use :cPyFrame_GetBuiltins.
    • f_code: use :cPyFrame_GetCode.
    • f_gen: use :cPyFrame_GetGenerator.
    • f_globals: use :cPyFrame_GetGlobals.
    • f_iblock: removed.
    • f_lasti: use :cPyFrame_GetLasti. Code using f_lasti with PyCode_Addr2Line() should use :cPyFrame_GetLineNumber instead; it may be faster.
    • f_lineno: use :cPyFrame_GetLineNumber
    • f_locals: use :cPyFrame_GetLocals.
    • f_stackdepth: removed.
    • f_state: no public API (renamed to f_frame.f_state).
    • f_trace: no public API.
    • f_trace_lines: use PyObject_GetAttrString((PyObject*)frame, "f_trace_lines").
    • f_trace_opcodes: use PyObject_GetAttrString((PyObject*)frame, "f_trace_opcodes").
    • f_localsplus: no public API (renamed to f_frame.localsplus).
    • f_valuestack: removed.

    The Python frame object is now created lazily. A side effect is that the f_back member must not be accessed directly, since its value is now also computed lazily. The :cPyFrame_GetBack function must be called instead.

    Debuggers that accessed the f_locals directly must call :cPyFrame_GetLocals instead. They no longer need to call :cPyFrame_FastToLocalsWithError or :cPyFrame_LocalsToFast, in fact they should not call those functions. The necessary updating of the frame is now managed by the virtual machine.

    Code defining PyFrame_GetCode() on Python 3.8 and older:

    #if PY_VERSION_HEX < 0x030900B1
static inline PyCodeObject* PyFrame_GetCode(PyFrameObject *frame)
{
    Py_INCREF(frame->f_code);
    return frame->f_code;
}
#endif

    Code defining PyFrame_GetBack() on Python 3.8 and older:

    #if PY_VERSION_HEX < 0x030900B1
static inline PyFrameObject* PyFrame_GetBack(PyFrameObject *frame)
{
    Py_XINCREF(frame->f_back);
    return frame->f_back;
}
#endif

    Or use the pythoncapi_compat project to get these two functions on older Python versions.

  • Changes of the :cPyThreadState structure members:

    • frame: removed, use :cPyThreadState_GetFrame (function added to Python 3.9 by 40429). Warning: the function returns a strong reference, need to call :cPy_XDECREF.
    • tracing: changed, use :cPyThreadState_EnterTracing and :cPyThreadState_LeaveTracing (functions added to Python 3.11 by 43760).
    • recursion_depth: removed, use (tstate->recursion_limit - tstate->recursion_remaining) instead.
    • stackcheck_counter: removed.

    Code defining PyThreadState_GetFrame() on Python 3.8 and older:

    #if PY_VERSION_HEX < 0x030900B1
static inline PyFrameObject* PyThreadState_GetFrame(PyThreadState *tstate)
{
    Py_XINCREF(tstate->frame);
    return tstate->frame;
}
#endif

    Code defining PyThreadState_EnterTracing() and PyThreadState_LeaveTracing() on Python 3.10 and older:

    #if PY_VERSION_HEX < 0x030B00A2
static inline void PyThreadState_EnterTracing(PyThreadState *tstate)
{
    tstate->tracing++;
#if PY_VERSION_HEX >= 0x030A00A1
    tstate->cframe->use_tracing = 0;
#else
    tstate->use_tracing = 0;
#endif
}

static inline void PyThreadState_LeaveTracing(PyThreadState *tstate)
{
    int use_tracing = (tstate->c_tracefunc != NULL || tstate->c_profilefunc != NULL);
    tstate->tracing--;
#if PY_VERSION_HEX >= 0x030A00A1
    tstate->cframe->use_tracing = use_tracing;
#else
    tstate->use_tracing = use_tracing;
#endif
}
#endif

    Or use the pythoncapi_compat project to get these functions on old Python functions.

  • Distributors are encouraged to build Python with the optimized Blake2 library libb2.

Deprecated

  • Deprecate the following functions to configure the Python initialization:

    • :cPySys_AddWarnOptionUnicode
    • :cPySys_AddWarnOption
    • :cPySys_AddXOption
    • :cPySys_HasWarnOptions
    • :cPy_SetPath
    • :cPy_SetProgramName
    • :cPy_SetPythonHome
    • :cPy_SetStandardStreamEncoding
    • :c_Py_SetProgramFullPath

    Use the new :cPyConfig API of the Python Initialization Configuration <init-config> instead (587). (Contributed by Victor Stinner in 44113.)

  • Deprecate the ob_shash member of the :cPyBytesObject. Use :cPyObject_Hash instead. (Contributed by Inada Naoki in 46864.)

Removed

  • :cPyFrame_BlockSetup and :cPyFrame_BlockPop have been removed. (Contributed by Mark Shannon in 40222.)

  • Remove the following math macros using the errno variable:

    • Py_ADJUST_ERANGE1()
    • Py_ADJUST_ERANGE2()
    • Py_OVERFLOWED()
    • Py_SET_ERANGE_IF_OVERFLOW()
    • Py_SET_ERRNO_ON_MATH_ERROR()

    (Contributed by Victor Stinner in 45412.)

  • Remove Py_UNICODE_COPY() and Py_UNICODE_FILL() macros, deprecated since Python 3.3. Use PyUnicode_CopyCharacters() or memcpy() (wchar_t* string), and PyUnicode_Fill() functions instead. (Contributed by Victor Stinner in 41123.)
  • Remove the pystrhex.h header file. It only contains private functions. C extensions should only include the main <Python.h> header file. (Contributed by Victor Stinner in 45434.)
  • Remove the Py_FORCE_DOUBLE() macro. It was used by the Py_IS_INFINITY() macro. (Contributed by Victor Stinner in 45440.)

  • The following items are no longer available when :cPy_LIMITED_API is defined:

    • :cPyMarshal_WriteLongToFile
    • :cPyMarshal_WriteObjectToFile
    • :cPyMarshal_ReadObjectFromString
    • :cPyMarshal_WriteObjectToString
    • the Py_MARSHAL_VERSION macro

    These are not part of the limited API <stable-abi-list>.

    (Contributed by Victor Stinner in 45474.)

  • Exclude :cPyWeakref_GET_OBJECT from the limited C API. It never worked since the :cPyWeakReference structure is opaque in the limited C API. (Contributed by Victor Stinner in 35134.)
  • Remove the PyHeapType_GET_MEMBERS() macro. It was exposed in the public C API by mistake, it must only be used by Python internally. Use the PyTypeObject.tp_members member instead. (Contributed by Victor Stinner in 40170.)
  • Remove the HAVE_PY_SET_53BIT_PRECISION macro (moved to the internal C API). (Contributed by Victor Stinner in 45412.)

  1. A similar optimization already existed since Python 3.8. 3.11 specializes for more forms and reduces some overhead.↩

  2. A similar optimization already existed since Python 3.10. 3.11 specializes for more forms. Furthermore, all attribute loads should be sped up by 45947.↩